KERYX
The archive← New dispatch
Archived dispatch

We are worried that a public benchmark is already in model training data. What practical checks can a small lab do without access to the training corpus?

Lowconfidence— Only source excerpts are delivered; complete synthesis and per-assertion support remain unverified. Evidence assessment: no citation passed the evidence gate

· llm:deepseek:deepseek-v4-pro

Review the original question, current payer, source cap and visibility before asking again. Reopening does not submit or spend anything. The saved report stays unchanged.

§ IIThe reading0 cited
Lowconfidence— Only source excerpts are delivered; complete synthesis and per-assertion support remain unverified. Evidence assessment: no citation passed the evidence gatedeep researchpreview plan 1/2 claimsSelected sources: 1/1 proposals

Recorded evidence

Targets with inspectable excerpts: 0/2.

Excerpt presence and estimated support do not establish a complete or factually correct answer.

Targets with missing excerpts or a low recorded support estimate: 2.

Inspect targets and prepare a follow-up

  1. What practical checks can a small lab perform to detect whether a public benchmark is already in model training data, given no access to the training corpus?

    No qualifying excerpt recorded · 0% estimated

    Prepare follow-up: What practical checks can a small lab perform to detect whether a public benchmark is already in model training data, given no access to the training corpus?

  2. What are the limitations of these checks in terms of reliability and false positives/negatives?

    No qualifying excerpt recorded · 0% estimated

    Prepare follow-up: What are the limitations of these checks in terms of reliability and false positives/negatives?

Opens an editable question. Review the question and budget before starting a new research run, which may incur charges.

No supported answer. The read content supplied no qualifying excerpts for the research targets; the draft is withheld as a conclusion.

Research target 1

Requested topic (unverified): “What practical checks can a small lab perform to detect whether a public benchmark is already in model training data, given no access to the training corpus?”

Evidence gap: no qualifying excerpt for this research target.

Research target 2

Requested topic (unverified): “What are the limitations of these checks in terms of reliability and false positives/negatives?”

Evidence gap: no qualifying excerpt for this research target.

Excerpts establish source grounding, not factual truth, entailment or whole-paper coverage. Support and coverage are estimates, not certification of a complete answer. Source statements may be wrong or conflicting. Draft conclusions are withheld; inspect the original text and obtain further review. Payment states remain in the separate receipt.

Evidence ledger — recorded source excerpts

Research targets are unverified topics. Coverage is an estimate of excerpt support, not proof of entailment, factual truth or a complete answer.

  1. Requested topic (unverified): “What practical checks can a small lab perform to detect whether a public benchmark is already in model training data, given no access to the training corpus?”

    0% estimated

    No qualifying excerpt recorded

  2. Requested topic (unverified): “What are the limitations of these checks in terms of reliability and false positives/negatives?”

    0% estimated

    No qualifying excerpt recorded

No inspectable non-demo excerpts are recorded for source inspection.

Research evidence matrix

Compare unverified research targets with cited sources and inspect recorded excerpts. An empty cell means no inspectable excerpt was recorded; it does not establish whether a claim is true, false, or disputed. Coverage and agent confidence do not prove entailment, measured accuracy or complete synthesis.

Inspect the evidence

Search target text, retained excerpts and article details.

Article and excerpt filters

Show targets with an excerpt from this exact citation. Other cited sources remain visible for comparison.

2 of 2 targets shown

Downloads include the complete evidence matrix, including targets hidden by these filters.

  1. Research target (unverified)

    What practical checks can a small lab perform to detect whether a public benchmark is already in model training data, given no access to the training corpus?

    Inspection status: No inspectable excerpt recorded

  2. Research target (unverified)

    What are the limitations of these checks in terms of reliability and false positives/negatives?

    Inspection status: No inspectable excerpt recorded

Helpful?

Loading feedback…

Spent
$0
To creators
—
Decisions
0 bought · 132 skipped
Public / cached reads
3 public reads
llm:deepseek:deepseek-v4-prolive on Arc mainnet
Decision log · 163 steps
§ IThe decision$0 settled / $0.05
0%

Scroll inside the log to read earlier steps. Jump to latest to follow new steps.

  1. Decompose

    Breaking down: "We are worried that a public benchmark is already in model training data. What practical checks can a small lab do without access to the training corpus?"

  2. Decompose

    Identified 2 research target(s) to investigate; these are not established facts

  3. Decompose

    Deep mode: up to 4 paid/cached/public reads plus one bounded gap-expansion pass when needed.

  4. Discover

    Article library searched across retained history: 28 matching article(s). Different articles from one publisher may support different research targets; preview matches alone are not evidence.

  5. Discover

    Web search: 3/3 planned queries attempted, 0 succeeded, 0 public page previews/leads, 3 unavailable queries. Previews and supplied URLs are discovery only. Public reads spend no USDC; model and service operating costs remain separate.

  6. Discover

    Discovered 3 verified creator candidate(s) from 2 publisher(s) and 130 free public reference candidate(s)

  7. Pre-check

    Claim-aware portfolio (exhaustive; bounded selection, not a claim of global optimality) selected 1/1 positive proposal(s): 1 free/cache selections + 0 paid fresh selections, predicting 1/2 claim(s) above the evidence floor with $0.000000/$0.025000 fetch USDC reserved.

  8. Pre-check

    Free-preview pre-check covers 1/2 sub-claims (50%). The agent may buy only claim-targeted sources and will label the answer provisional if paid evidence stays thin.

  9. DecideCACHE
    MIT News: Artificial Intelligence - When AI art has no author: Study finds generated images often can’t be traced to training data$0 · EV 60%

    MIT News article discusses tracing generated images to training data, which may offer insights into limitations of detecting training data inclusion, relevant to target 1. - free public feed reference; no purchase or creator reward. — selected for the claim-aware evidence portfolio (targets claim 2; 0 fetch USDC, 1 attention slot).

  10. DecideSKIP
    Apache Software Foundation - The Apache® Software Foundation Announces Keynote Speakers for Community Over Code Glasgow 2026$0 · EV 0%

    Announcement about conference keynotes; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  11. DecideSKIP
    Apache Software Foundation - The Apache® Software Foundation Announces New Top-Level Projects$0 · EV 0%

    Announcement about new top-level projects; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  12. DecideSKIP
    Apache Software Foundation - Security scanning at Foundation scale$0 · EV 0%

    Security scanning at Foundation scale; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  13. DecideSKIP
    Apache Arrow - Apache Arrow 24.0.0 Release$0 · EV 0%

    Apache Arrow release notes; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  14. DecideSKIP
    Apache Arrow - Apache Arrow 25.0.0 Release$0 · EV 0%

    Apache Arrow release notes; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  15. DecideSKIP
    Apache Arrow - Apache Arrow 26.0.0 Release$0 · EV 0%

    Apache Arrow release notes; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  16. DecideSKIP
    arXiv - Fair Moderation, Equitable Access, and AI: arXiv’s Updated Rate Limit Policy$0 · EV 0%

    arXiv rate limit policy; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  17. DecideSKIP
    arXiv - Modernizing arXiv technology: Updates on the arXiv-CE project$0 · EV 0%

    arXiv technology modernization; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  18. DecideSKIP
    arXiv cs.AI paper announcements - MSP: Modality Self-Play for Training Multimodal LLMs Without Paired Cross-Modal Data$0 · EV 0%

    Paper on multimodal self-play; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  19. DecideSKIP
    arXiv cs.AI paper announcements - Can LLMs Tell Stories Out of Order? A Corpus and Benchmark for Nonlinear Narrative Generation$0 · EV 0%

    Paper on nonlinear narrative generation; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  20. DecideSKIP
    arXiv cs.AI paper announcements - A Jury, Not an Electorate: Label-Free Reliability Estimation for Ensembles of Language Models$0 · EV 0%

    Paper on ensemble reliability estimation; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  21. DecideSKIP
    Bank for International Settlements - High public debt, NBFIs and the risk of market dysfunction$0 · EV 0%

    BIS bulletin on public debt and NBFIs; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  22. DecideSKIP
    Bank for International Settlements - Offshore forwards and exchange rate transmission in emerging markets$0 · EV 0%

    BIS working paper on offshore forwards; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  23. DecideSKIP
    Chip Huyen - Predictive Human Preference: From Model Ranking to Model Routing$0 · EV 0%

    Article on predictive human preference; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  24. DecideSKIP
    Chip Huyen - Common pitfalls when building generative AI applications$0 · EV 0%

    Article on common pitfalls in generative AI; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  25. DecideSKIP
    Chrome for Developers - Connection allowlists: Secure your web application's network access$0 · EV 0%

    Chrome connection allowlists; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  26. DecideSKIP
    Chrome for Developers - What's New in WebGPU (Chrome 155-156)$0 · EV 0%

    WebGPU updates; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  27. DecideSKIP
    Cloudflare Workers - Secure all your internal vibe-coded applications — in one click$0 · EV 0%

    Cloudflare Workers security for vibe-coded apps; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  28. DecideSKIP
    Cloudflare Workers - Detect and send production issues straight to your agent$0 · EV 0%

    Cloudflare Workers production issue detection; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  29. DecideSKIP
    Cloudflare Workers - A fully programmable secure web gateway$0 · EV 0%

    Article about programmable secure web gateway, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  30. DecideSKIP
    Creative Commons - Guidance for Indigenous Collections and Data$0 · EV 0%

    Guidance for Indigenous Collections and Data, unrelated to ML benchmark contamination. - free public feed reference; no purchase or creator reward.

  31. DecideSKIP
    Creative Commons - A Principled Approach as the Sands Shift$0 · EV 0%

    Creative Commons AI principles, not practical checks for benchmark contamination. - free public feed reference; no purchase or creator reward.

  32. DecideSKIP
    Google DeepMind - Google DeepMind and Singapore Government collaborate to build fine-tuned AI model for scam detection$0 · EV 0%

    Scam detection model collaboration, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  33. DecideSKIP
    Google DeepMind - Accelerating the search for antimicrobial peptides with Google Antigravity$0 · EV 0%

    Antimicrobial peptide discovery, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  34. DecideSKIP
    Distill - Weight Banding$0 · EV 0%

    Weight banding in visual models, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  35. DecideSKIP
    Distill - A Gentle Introduction to Graph Neural Networks$0 · EV 0%

    Graph neural networks introduction, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  36. DecideSKIP
    Django - Call for volunteers: Fundraising Working Group$0 · EV 0%

    Django fundraising working group, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  37. DecideSKIP
    Django - Django security reporting update$0 · EV 0%

    Django security reporting update, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  38. DecideSKIP
    Django - Django security releases issued: 6.1.2, 6.0.9, and 5.2.18$0 · EV 0%

    Django security releases, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  39. DecideSKIP
    Directory of Open Access Journals - Sustaining trusted open access infrastructure: Thank you to our publisher and data user supporters$0 · EV 0%

    DOAJ infrastructure support, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  40. DecideSKIP
    Directory of Open Access Journals - From partnership to policy: How DOAJ and Ukraine are strengthening open access publishing$0 · EV 0%

    DOAJ Ukraine partnership, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  41. DecideSKIP
    Docker - Docker Hub security advisory: personal access tokens transferred during account-to-organization conversions$0 · EV 0%

    Docker Hub security advisory, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  42. DecideSKIP
    Docker - AI Agent Security Risks: What Can Go Wrong When Agents Act on Their Own$0 · EV 0%

    AI agent security risks, not about benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  43. DecideSKIP
    Docker - From Containers to Sandboxes: Developer Trust in the Agentic Era$0 · EV 0%

    Containers to sandboxes trust, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  44. DecideSKIP
    DuckDB - A DuckDB Database with No Data in It$0 · EV 0%

    DuckDB database with no data, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  45. DecideSKIP
    DuckDB - Agent Mode in the DuckDB CLI$0 · EV 0%

    DuckDB agent mode, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  46. DecideSKIP
    Eugene Yan - Training an LLM-RecSys Hybrid for Steerable Recs with Semantic IDs$0 · EV 0%

    LLM-RecSys hybrid, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  47. DecideSKIP
    Eugene Yan - Patterns for Building Cybersecurity Evals$0 · EV 0%

    Cybersecurity evals patterns, not about benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  48. DecideSKIP
    GitHub Engineering - Staying Evergreen: How we automated secure access and repo hygiene in a shared org$0 · EV 0%

    GitHub repo hygiene automation, unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  49. DecideSKIP
    GitHub Engineering - Your alt text passes automated checks. That doesn’t mean it’s any good.$0 · EV 0%

    Article about alt text quality on web pages; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  50. DecideSKIP
    The Go Blog - Arch-specific SIMD in Go$0 · EV 0%

    Go blog post about SIMD APIs; no relevance to detecting benchmark contamination. - free public feed reference; no purchase or creator reward.

  51. DecideSKIP
    The Go Blog - Platform-independent SIMD in Go$0 · EV 0%

    Go blog post about SIMD; unrelated to benchmark contamination checks. - free public feed reference; no purchase or creator reward.

  52. DecideSKIP
    The Go Blog - Interactive refactoring in gopls$0 · EV 0%

    Go blog post about interactive refactoring; no relevance to benchmark contamination. - free public feed reference; no purchase or creator reward.

  53. DecideSKIP
    Google Research - Research@ 2026: 10 ways we're amplifying human ingenuity$0 · EV 0%

    Google Research overview of amplifying human ingenuity; not about benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  54. DecideSKIP
    Google Research - Unlocking Earth AI’s planetary geospatial foundation models for global public health$0 · EV 0%

    Article about geospatial foundation models for public health; unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  55. DecideSKIP
    Google Research - Grounding AI agents in biological reality: Building blocks for autonomous wet labs$0 · EV 0%

    Article about AI agents for wet labs; no relevance to benchmark contamination checks. - free public feed reference; no purchase or creator reward.

  56. DecideSKIP
    IPFS - New Stewards for Public IPFS Network Services$0 · EV 0%

    IPFS network services stewardship; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  57. DecideSKIP
    Kotlin - What Makes a Great KotlinConf Proposal?$0 · EV 0%

    KotlinConf proposal guidance; no relevance to benchmark contamination. - free public feed reference; no purchase or creator reward.

  58. DecideSKIP
    Kotlin - Google Summer of Code 2026 With the Kotlin Foundation: Recap$0 · EV 0%

    Kotlin GSoC recap; unrelated to benchmark contamination checks. - free public feed reference; no purchase or creator reward.

  59. DecideSKIP
    Kotlin - A More Reliable Compilation Scheme for Kotlin Multiplatform Modules$0 · EV 0%

    Kotlin compilation scheme; no relevance to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  60. DecideSKIP
    Krebs on Security - Data Broker Radaris Loses Domains in Privacy Fight$0 · EV 0%

    Krebs article about data broker domain loss; unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  61. DecideSKIP
    Krebs on Security - Feds Suspect Ransom Negotiator Ran ShinyHunters Extortion Site$0 · EV 0%

    Krebs article about ShinyHunters extortion; no relevance to benchmark contamination checks. - free public feed reference; no purchase or creator reward.

  62. DecideSKIP
    Lilian Weng - Thinking about High-Quality Human Data$0 · EV 0%

    Lilian Weng post about high-quality human data; not about detecting benchmark contamination. - free public feed reference; no purchase or creator reward.

  63. DecideSKIP
    Lilian Weng - Harness Engineering for Self-Improvement$0 · EV 0%

    Lilian Weng post about harness engineering for self-improvement; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  64. DecideSKIP
    Martin Fowler - Fragments: September 24$0 · EV 0%

    Martin Fowler fragments about AI risks; no relevance to benchmark contamination checks. - free public feed reference; no purchase or creator reward.

  65. DecideSKIP
    Martin Fowler - Are Harnesses Worth the Cost?$0 · EV 0%

    Martin Fowler article about harnesses for coding agents; not about benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  66. DecideSKIP
    Martin Fowler - Bliki: Sensible Default$0 · EV 0%

    Martin Fowler bliki about sensible defaults; unrelated to benchmark contamination. - free public feed reference; no purchase or creator reward.

  67. DecideSKIP
    MDN Blog - Image formats: Pixel data from encoders to decoders$0 · EV 0%

    MDN article about image formats; no relevance to benchmark contamination checks. - free public feed reference; no purchase or creator reward.

  68. DecideSKIP
    MDN Blog - Introducing the MDN MCP server$0 · EV 0%

    MDN article about MCP server; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  69. DecideSKIP
    Microsoft Engineering - How we built the Microsoft Learn MCP Server$0 · EV 0%

    Article about Microsoft Learn MCP Server; unrelated to detecting benchmark contamination in model training data. - free public feed reference; no purchase or creator reward.

  70. DecideSKIP
    Microsoft Engineering - Engineering and algorithmic interventions for multimodal post-training at Microsoft scale$0 · EV 0%

    Article about multimodal post-training interventions; not about detecting whether a benchmark is in training data. - free public feed reference; no purchase or creator reward.

  71. DecideSKIP
    Microsoft Engineering - Common annotated security keys$0 · EV 0%

    Article about annotated security keys; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  72. DecideSKIP
    Microsoft Research - Agent Lightning v1.0: A 3,500-Line Lightweight Agentic RL Framework for Training Agents with Real Harnesses$0 · EV 0%

    Article about Agent Lightning RL framework; not about detecting benchmark leakage. - free public feed reference; no purchase or creator reward.

  73. DecideSKIP
    Microsoft Research - One year in: How Microsoft Research Asia – Singapore is advancing research, partnership and talent for real-world impact$0 · EV 0%

    Article about Microsoft Research Asia Singapore; unrelated to benchmark contamination checks. - free public feed reference; no purchase or creator reward.

  74. DecideSKIP
    MIT News: Artificial Intelligence - Giving AI a “crystal clear” view of the materials world$0 · EV 0%

    Article about AI for materials science; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  75. DecideSKIP
    Mozilla AI - Mistral Large 4 is live in Otari$0 · EV 0%

    Article about Mistral Large 4 in Otari; unrelated to detecting benchmark leakage. - free public feed reference; no purchase or creator reward.

  76. DecideSKIP
    Mozilla AI - Building the Missing Half$0 · EV 0%

    Article about building infrastructure for open models; not about benchmark contamination checks. - free public feed reference; no purchase or creator reward.

  77. DecideSKIP
    Mozilla Hacks - Intent to Ship: JPEG XL$0 · EV 0%

    Article about JPEG XL browser support; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  78. DecideSKIP
    Mozilla Hacks - PACT: Anonymous Credentials for the Web$0 · EV 0%

    Article about anonymous credentials for the web; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  79. DecideSKIP
    Netlify - Choosing an AI model: one prompt, 11 models, very different results$0 · EV 0%

    Article about choosing AI models for builds; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  80. DecideSKIP
    Netlify - 5x faster Edge Functions: v8 isolates to Unikraft MicroVMs$0 · EV 0%

    Article about Edge Functions performance; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  81. DecideSKIP
    Ollama - Reflection's Beam is freely available on Ollama for a limited time$0 · EV 0%

    Article about Reflection's Beam model; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  82. DecideSKIP
    OpenAI Developers - OpenAI Developers plugin$0 · EV 0%

    Article about OpenAI Developers plugin; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  83. DecideSKIP
    OpenAI Developers - API deployment checklist$0 · EV 0%

    Article about API deployment checklist; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  84. DecideSKIP
    OpenAlex - When affiliation errors become a research security problem$0 · EV 0%

    Article about affiliation errors in research security; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  85. DecideSKIP
    OpenAlex - Trusted journal lists come to OpenAlex$0 · EV 0%

    Article about trusted journal lists; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  86. DecideSKIP
    Open Source Security Foundation - OpenSSF Newsletter – September 2026$0 · EV 0%

    OpenSSF newsletter; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  87. DecideSKIP
    Open Source Security Foundation - What’s in the SOSS? Podcast #75 – S3E27 From Upstream to Downstream: Managing Open Source Risk in a Changing Regulatory Era with Vincent Danen$0 · EV 0%

    Podcast about open source risk management; unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  88. DecideSKIP
    Open Source Security Foundation - Case Study: GitHub Expands Malware Advisories Across Eight Ecosystems with OpenSSF$0 · EV 0%

    Topic is GitHub malware advisories and supply-chain security, unrelated to detecting benchmark contamination in model training data. - free public feed reference; no purchase or creator reward.

  89. DecideSKIP
    Our World in Data - The United States is the world’s largest oil producer$0 · EV 0%

    Topic is oil production statistics, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  90. DecideSKIP
    Our World in Data - How many people in the world identify with a particular religion?$0 · EV 0%

    Topic is global religious affiliation statistics, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  91. DecideSKIP
    Our World in Data - In these nine African countries, average incomes have more than doubled since 1990$0 · EV 0%

    Topic is African economic growth statistics, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  92. DecideSKIP
    PLOS - PLOS responds to Office of Management and Budget (OMB) proposed rule$0 · EV 0%

    Topic is PLOS response to OMB proposed rule on federal research funding, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  93. DecideSKIP
    PLOS - PLOS launches two journals to address critical real-world challenges$0 · EV 0%

    Topic is PLOS journal launches, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  94. DecideSKIP
    PLOS - Renewing our partnership to advance open science across Africa$0 · EV 0%

    Topic is open science partnership in Africa, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  95. DecideSKIP
    PostgreSQL - Dasha 1.8: index recommendations, I/O analysis, schema checks and log insights$0 · EV 0%

    Topic is PostgreSQL performance dashboard Dasha, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  96. DecideSKIP
    Protocol Labs - Data Sovereignty in a Privacy First Future$0 · EV 0%

    Topic is data sovereignty and privacy, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  97. DecideSKIP
    Protocol Labs - IPTS Acquires Talent Protocol’s Record of 6.6 Million Builders$0 · EV 0%

    Topic is IPTS acquisition of Talent Protocol, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  98. DecideSKIP
    Python Blog - Free-Threaded Python Post-Era (Python Language Summit 2026)$0 · EV 0%

    Topic is free-threaded Python concurrency model, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  99. DecideSKIP
    Python Blog - Python 3.16.0 alpha 1 is now available$0 · EV 0%

    Topic is Python 3.16 alpha release, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  100. DecideSKIP
    PyTorch - Bringing Elastically Deterministic Data Loading to TorchTitan with Zephon$0 · EV 0%

    Topic is deterministic data loading in TorchTitan, unrelated to detecting benchmark contamination in training data. - free public feed reference; no purchase or creator reward.

  101. DecideSKIP
    PyTorch - Bringing ExecuTorch to Arm Cortex-M CPUs$0 · EV 0%

    Topic is ExecuTorch on Arm Cortex-M CPUs, unrelated to detecting benchmark contamination in training data. - free public feed reference; no purchase or creator reward.

  102. DecideSKIP
    React Blog - Critical Security Vulnerability in React Server Components$0 · EV 0%

    Topic is React Server Components security vulnerability, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  103. DecideSKIP
    React Blog - React 19.3$0 · EV 0%

    Topic is React 19.3 release features, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  104. DecideSKIP
    Rust Blog - Rust debugging survey 2026 results$0 · EV 0%

    Topic is Rust debugging survey results, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  105. DecideSKIP
    Rust Blog - Announcing Rust 1.99.1$0 · EV 0%

    Topic is Rust 1.99.1 release notes, unrelated to model training data contamination checks. - free public feed reference; no purchase or creator reward.

  106. DecideSKIP
    Schneier on Security - Backdooring Open-Weight AI Models$0 · EV 0%

    Topic is backdooring open-weight AI models, not detecting benchmark contamination in training data. - free public feed reference; no purchase or creator reward.

  107. DecideSKIP
    Sebastian Raschka - Focusing on Post-Training$0 · EV 0%

    Topic is post-training investment, unrelated to detecting benchmark contamination in training data. - free public feed reference; no purchase or creator reward.

  108. DecideSKIP
    Sebastian Raschka - Pacing != Pacing Development$0 · EV 0%

    Topic is AI model pacing framework, unrelated to detecting benchmark contamination in training data. - free public feed reference; no purchase or creator reward.

  109. DecideSKIP
    Spotify Engineering - AI Changed How Spotify Builds. What We Learned (and Fixed) About Quality at Higher Velocity$0 · EV 0%

    Article about AI quality and velocity at Spotify, unrelated to detecting benchmark contamination in training data. - free public feed reference; no purchase or creator reward.

  110. DecideSKIP
    Spotify Engineering - Advancing Generative Recommendation Inference: our collaboration with NVIDIA on TensorRT LLM$0 · EV 0%

    Article about generative recommendation inference with NVIDIA, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  111. DecideSKIP
    Stellar Blog - Practical Confidential Stablecoins: An Issuer-Controlled Architecture$0 · EV 0%

    Article about confidential stablecoins, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  112. DecideSKIP
    Stellar Blog - No one to call$0 · EV 0%

    Article about blockchain counterparties, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  113. DecideSKIP
    Stripe Blog - Sync real-time Stripe data to Postgres with Stripe Data Pipeline$0 · EV 0%

    Article about Stripe Data Pipeline, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  114. DecideSKIP
    Stripe Blog - New ways for fintechs to move money and expand globally$0 · EV 0%

    Article about fintech money movement, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  115. DecideSKIP
    Supabase - How we made the Supabase Select 2026 badge$0 · EV 0%

    Article about a conference badge, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  116. DecideSKIP
    Supabase - Sync real-time Stripe data to Supabase with Stripe Data Pipeline$0 · EV 0%

    Article about Stripe data sync, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  117. DecideSKIP
    Supabase - Supabase Storage: major performance, security, and reliability updates$0 · EV 0%

    Article about Supabase Storage updates, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  118. DecideSKIP
    Super Simple Songs - Kids Songs - When We Play Hockey 🏒 | Super Simple Songs & ‪@NHL Hockey Song for Kids$0 · EV 0%

    Children's hockey song video, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  119. DecideSKIP
    Super Simple Songs - Kids Songs - 🔴 Trick or Treat With Super Simple Live! 🎃 | Best Halloween Kids Songs | Super Simple Songs$0 · EV 0%

    Halloween children's song video, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  120. DecideSKIP
    Tailscale - Tailcat: Tailscale without Tailscale, by Tailscale$0 · EV 0%

    Article about Tailcat networking tool, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  121. DecideSKIP
    Tailscale - It's time to make your own RSS reader$0 · EV 0%

    Article about self-hosted RSS readers, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  122. DecideSKIP
    V8 - How we made JSON.stringify more than twice as fast$0 · EV 0%

    Article about JSON.stringify performance, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  123. DecideSKIP
    V8 - WebAssembly JSPI is going to origin trial$0 · EV 0%

    Article about WebAssembly JSPI, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  124. DecideSKIP
    Vicki Boykis - We should be more tired than the model$0 · EV 0%

    Article about cognitive effects of AI coding, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  125. DecideSKIP
    Vicki Boykis - Running local models is good now$0 · EV 0%

    Article about running local models, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  126. DecideSKIP
    vLLM - Finish RL Training Faster: Bitwise-Consistent, High-Performance, Agent-Driven$0 · EV 0%

    Article about RL training performance, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  127. DecideSKIP
    vLLM - Serving Reflection Beam on vLLM from Day 0$0 · EV 0%

    Article about serving Reflection Beam, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  128. DecideSKIP
    W3C Blog - We’re on a roll - EPUB 3.4 is coming$0 · EV 0%

    Article about EPUB 3.4, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  129. DecideSKIP
    W3C Blog - Introducing group activity pages$0 · EV 0%

    Article about W3C group activity pages, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  130. DecideSKIP
    web.dev - Find features to support for Interop 2027 with your coding agent$0 · EV 0%

    Article about Interop 2027 proposals, unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  131. DecideSKIP
    web.dev - New to the web platform in September$0 · EV 0%

    Preview is about new web platform features in September 2026, unrelated to detecting benchmark contamination in model training data. - free public feed reference; no purchase or creator reward.

  132. DecideSKIP
    WebKit - Release Notes for Safari Technology Preview 252$0 · EV 0%

    Safari Technology Preview release notes focus on browser accessibility fixes, not on benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  133. DecideSKIP
    WebKit - Release Notes for Safari Technology Preview 254$0 · EV 0%

    Safari Technology Preview release notes focus on browser accessibility fixes, not on benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  134. DecideSKIP
    Wikimedia Diff - Beyond Wikipedia: Training Editors with Wikimedia Nigeria, Rivers Network$0 · EV 0%

    Wikimedia Diff article about training editors in Nigeria is unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  135. DecideSKIP
    Wikimedia Diff - Bridges Built: EduWiki Hub and ESEAP Hub Strengthen Access to Wikimedia Education$0 · EV 0%

    Wikimedia Diff article about education hubs is unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  136. DecideSKIP
    Wikimedia Diff - Building Bridges between Bangladesh-China through Wikimedia$0 · EV 0%

    Wikimedia Diff article about Bangladesh-China collaboration is unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  137. DecideSKIP
    x402 - x402: An open standard for internet-native payments$0 · EV 0%

    x402 payment standard article is unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  138. DecideSKIP
    x402 - Linux Foundation Announces Operational Launch of x402 Foundation to Standardize Internet-Native Payments for AI Agents and Applications$0 · EV 0%

    x402 Foundation launch announcement is unrelated to benchmark contamination detection. - free public feed reference; no purchase or creator reward.

  139. DecideSKIP
    Keryx owner release notes — first-party validation — Keryx v0.27.78$0.005 · EV 0%

    Keryx release notes are about software fixes, unrelated to benchmark contamination detection.

  140. DecideSKIP
    Keryx owner release notes — first-party validation — Keryx v0.27.76$0.005 · EV 0%

    Keryx release notes are about software fixes, unrelated to benchmark contamination detection.

  141. DecideSKIP
    Keryx Engineering (first-party) — How Keryx pays cited creators$0.002 · EV 0%

    Keryx engineering documentation about citation rewards is unrelated to benchmark contamination detection.

  142. Fetch

    Read MIT News: Artificial Intelligence - When AI art has no author: Study finds generated images often can’t be traced to training data - free public feed reference, no creator payment - S1

  143. Sufficiency

    Sub-claim "What practical checks can a small lab perform to detect whet…": 0% covered

  144. Sufficiency

    Sub-claim "What are the limitations of these checks in terms of reliabi…": 0% covered

  145. Sufficiency

    The gathered source S1 discusses attribution decay in generative models, showing that individual training examples often cannot be traced to outputs. It does not provide practical checks for detecting whether a benchmark is in training data, nor does it discuss limitations of such checks. Therefore, both sub-claims are unsupported. The assessment does not establish a complete supported answer for every requested part.

  146. Re-evaluate

    Sub-claim "What practical checks can a small lab perform to detect whet…": 0% covered — The gathered source (S1) discusses attribution decay in generative image models and the difficulty of tracing outputs to training data, but it does not describe any practical checks a small lab could perform to detect whether a public benchmark was included in training data. No procedures, tests, or methods are provided.

  147. Re-evaluate

    Sub-claim "What are the limitations of these checks in terms of reliabi…": 0% covered — No checks are described in the gathered content, so there is no discussion of their limitations, reliability, or false positive/negative rates. The source only mentions that attribution may be impossible at scale, which is not a limitation of a specific detection check.

  148. Re-evaluate

    Both sub-claims are completely uncovered (coverage 0.0). The remaining budget is $0.025, and all skipped sources are priced at $0.00 or $0.005, so affordable sources can be purchased. The most promising candidates are 'Schneier on Security - Backdooring Open-Weight AI Models' (public:schneier), which may discuss model manipulation and detection, and 'Chip Huyen - Common pitfalls when building generative AI applications' (public:chip-huyen:article:006d0f5af8f9abfeaf302a2e167701ccc4d8de10c59581c8a1caa1c1975cc0ed), which may cover benchmark leakage or evaluation pitfalls. These are prioritized based on topical relevance to AI model training and evaluation.

  149. Re-evaluate

    Filling gap - free public feed reference Schneier on Security, no creator payment - S2

  150. Re-evaluate

    Filling gap - free public feed reference Chip Huyen, no creator payment - S3

  151. Sufficiency

    Final check — "What practical checks can a small lab perform to detect whet…": 0% assessed

  152. Sufficiency

    Final check — "What are the limitations of these checks in terms of reliabi…": 0% assessed

  153. Sufficiency

    Final coverage assessment — S1 discusses attribution decay in generative models, showing that individual training examples become less attributable as dataset size grows, but it does not provide practical checks for detecting benchmark contamination. S2 covers backdooring open-weight models, unrelated to benchmark contamination. S3 is a general list of AI engineering pitfalls, mentioning training data lineage as a compliance issue but offering no detection methods. None of the sources provide practical checks or limitations for detecting whether a public benchmark is in training data without corpus access. The assessment does not establish a complete supported answer for every requested part.

  154. Synthesize

    Synthesizing a grounded answer from 3 source(s)…

  155. Evidence

    Relevance review returned; only checked excerpts can retain support, and review cannot raise it.

  156. Evidence

    Delivering qualified source excerpts; complete synthesis and per-assertion support remain unverified.

  157. Evidence

    Below support/reward gate — S1, research target 1, proposed support 20% (estimate, not entailment): “The scientists identified a phenomenon they call attribution decay, where the more data a generative model is trained on, the less any indiv…”

  158. Evidence

    Below support/reward gate — S1, research target 2, proposed support 20% (estimate, not entailment): “The scientists identified a phenomenon they call attribution decay, where the more data a generative model is trained on, the less any indiv…”

  159. Evidence

    Rejected 0 invalid evidence span(s) and 1 unsupported citation marker(s); rejected markers cannot receive citation rewards.

  160. Evidence

    No citation passed the evidence gate — the $0.025000 citation pool stays unspent; settled access tolls still stand.

  161. Synthesize

    Prepared source excerpts citing 0 source(s); complete synthesis is unverified

  162. Verdict

    Confidence: Low — Only source excerpts are delivered; complete synthesis and per-assertion support remain unverified. Evidence assessment: no citation passed the evidence gate.

  163. Done

    Done. Spent $0 across 0 confirmed/simulated payment(s) to creators.

Following latest steps

Sign in with the original report owner to review its decisions.

Read checkpoints

Verify locally replays supported recorded query-local reserve, release, debit and hold decisions, attribution normalization and exact micro-USDC citation/author allocations. Model proposals and eligibility are asserted inputs. Query capacity is separate from signer exposure; verification does not establish funding, reward authority or settlement.

Economic families or records marked unavailable are not verified. Missing, fractional-micro or unsafe operands are never repaired; an empty family means no captured economic decision, not complete monetary coverage.

Query-local budget records
Recorded decisions can be replayed
Citation allocation records
Not captured
Author split records
Not captured

Model proposals and selected plan actions are recorded when available. Model generation is not replayed or expected to reproduce; verification checks only deterministic checkpoints.

Micro-USDC is shown only where the existing exact converter accepts the amount. Other monetary assertions are unavailable; numeric predicate replay preserves the recorded comparison.

Recorded checkpoint policy
ordinary-query-economics-v4
SHA-256 retained by this report
3f391d7d16724f663f31540c9e0cddff2fb9eb8ad15922b9b6e209fc856ab0e6

Keep this digest separately from the downloaded sidecar. The offline audit command uses it as the expected digest.

Model proposal → selected plan → code checkpoint

BUY0
SKIP130
CACHE0
FREE3
ESCALATE0
UNKNOWN0

Latest terminal checkpoint outcome per recorded candidate; continuing checks and global stops are excluded. These are not settled purchases or quality scores.

Global stop checkpoints: 0 · Candidate rounds with continuing checks only: 0

CandidateRead roundModel proposedSelected planCheckpointCode resultReason / rule
650CACHECACHEselectionCONTINUEselected-plan
10SKIPSKIPselectionSKIPplan-skip
20SKIPSKIPselectionSKIPplan-skip
30SKIPSKIPselectionSKIPplan-skip
40SKIPSKIPselectionSKIPplan-skip
50SKIPSKIPselectionSKIPplan-skip
60SKIPSKIPselectionSKIPplan-skip
70SKIPSKIPselectionSKIPplan-skip
80SKIPSKIPselectionSKIPplan-skip
90SKIPSKIPselectionSKIPplan-skip
100SKIPSKIPselectionSKIPplan-skip
110SKIPSKIPselectionSKIPplan-skip
120SKIPSKIPselectionSKIPplan-skip
130SKIPSKIPselectionSKIPplan-skip
140SKIPSKIPselectionSKIPplan-skip
150SKIPSKIPselectionSKIPplan-skip
160SKIPSKIPselectionSKIPplan-skip
170SKIPSKIPselectionSKIPplan-skip
180SKIPSKIPselectionSKIPplan-skip
190SKIPSKIPselectionSKIPplan-skip
200SKIPSKIPselectionSKIPplan-skip
210SKIPSKIPselectionSKIPplan-skip
220SKIPSKIPselectionSKIPplan-skip
230SKIPSKIPselectionSKIPplan-skip
240SKIPSKIPselectionSKIPplan-skip
250SKIPSKIPselectionSKIPplan-skip
260SKIPSKIPselectionSKIPplan-skip
270SKIPSKIPselectionSKIPplan-skip
280SKIPSKIPselectionSKIPplan-skip
290SKIPSKIPselectionSKIPplan-skip
300SKIPSKIPselectionSKIPplan-skip
310SKIPSKIPselectionSKIPplan-skip
320SKIPSKIPselectionSKIPplan-skip
330SKIPSKIPselectionSKIPplan-skip
340SKIPSKIPselectionSKIPplan-skip
350SKIPSKIPselectionSKIPplan-skip
360SKIPSKIPselectionSKIPplan-skip
370SKIPSKIPselectionSKIPplan-skip
380SKIPSKIPselectionSKIPplan-skip
390SKIPSKIPselectionSKIPplan-skip
400SKIPSKIPselectionSKIPplan-skip
410SKIPSKIPselectionSKIPplan-skip
420SKIPSKIPselectionSKIPplan-skip
430SKIPSKIPselectionSKIPplan-skip
440SKIPSKIPselectionSKIPplan-skip
450SKIPSKIPselectionSKIPplan-skip
460SKIPSKIPselectionSKIPplan-skip
470SKIPSKIPselectionSKIPplan-skip
480SKIPSKIPselectionSKIPplan-skip
490SKIPSKIPselectionSKIPplan-skip
500SKIPSKIPselectionSKIPplan-skip
510SKIPSKIPselectionSKIPplan-skip
520SKIPSKIPselectionSKIPplan-skip
530SKIPSKIPselectionSKIPplan-skip
540SKIPSKIPselectionSKIPplan-skip
550SKIPSKIPselectionSKIPplan-skip
560SKIPSKIPselectionSKIPplan-skip
570SKIPSKIPselectionSKIPplan-skip
580SKIPSKIPselectionSKIPplan-skip
590SKIPSKIPselectionSKIPplan-skip
600SKIPSKIPselectionSKIPplan-skip
610SKIPSKIPselectionSKIPplan-skip
620SKIPSKIPselectionSKIPplan-skip
630SKIPSKIPselectionSKIPplan-skip
640SKIPSKIPselectionSKIPplan-skip
660SKIPSKIPselectionSKIPplan-skip
670SKIPSKIPselectionSKIPplan-skip
680SKIPSKIPselectionSKIPplan-skip
690SKIPSKIPselectionSKIPplan-skip
700SKIPSKIPselectionSKIPplan-skip
710SKIPSKIPselectionSKIPplan-skip
720SKIPSKIPselectionSKIPplan-skip
730SKIPSKIPselectionSKIPplan-skip
740SKIPSKIPselectionSKIPplan-skip
750SKIPSKIPselectionSKIPplan-skip
760SKIPSKIPselectionSKIPplan-skip
770SKIPSKIPselectionSKIPplan-skip
780SKIPSKIPselectionSKIPplan-skip
790SKIPSKIPselectionSKIPplan-skip
800SKIPSKIPselectionSKIPplan-skip
810SKIPSKIPselectionSKIPplan-skip
820SKIPSKIPselectionSKIPplan-skip
830SKIPSKIPselectionSKIPplan-skip
840SKIPSKIPselectionSKIPplan-skip
850SKIPSKIPselectionSKIPplan-skip
860SKIPSKIPselectionSKIPplan-skip
870SKIPSKIPselectionSKIPplan-skip
880SKIPSKIPselectionSKIPplan-skip
890SKIPSKIPselectionSKIPplan-skip
900SKIPSKIPselectionSKIPplan-skip
910SKIPSKIPselectionSKIPplan-skip
920SKIPSKIPselectionSKIPplan-skip
930SKIPSKIPselectionSKIPplan-skip
940SKIPSKIPselectionSKIPplan-skip
950SKIPSKIPselectionSKIPplan-skip
960SKIPSKIPselectionSKIPplan-skip
970SKIPSKIPselectionSKIPplan-skip
980SKIPSKIPselectionSKIPplan-skip
990SKIPSKIPselectionSKIPplan-skip
1000SKIPSKIPselectionSKIPplan-skip
1010SKIPSKIPselectionSKIPplan-skip
1020SKIPSKIPselectionSKIPplan-skip
1030SKIPSKIPselectionSKIPplan-skip
1040SKIPSKIPselectionSKIPplan-skip
1050SKIPSKIPselectionSKIPplan-skip
1060SKIPSKIPselectionSKIPplan-skip
1070SKIPSKIPselectionSKIPplan-skip
1080SKIPSKIPselectionSKIPplan-skip
1090SKIPSKIPselectionSKIPplan-skip
1100SKIPSKIPselectionSKIPplan-skip
1110SKIPSKIPselectionSKIPplan-skip
1120SKIPSKIPselectionSKIPplan-skip
1130SKIPSKIPselectionSKIPplan-skip
1140SKIPSKIPselectionSKIPplan-skip
1150SKIPSKIPselectionSKIPplan-skip
1160SKIPSKIPselectionSKIPplan-skip
1170SKIPSKIPselectionSKIPplan-skip
1180SKIPSKIPselectionSKIPplan-skip
1190SKIPSKIPselectionSKIPplan-skip
1200SKIPSKIPselectionSKIPplan-skip
1210SKIPSKIPselectionSKIPplan-skip
1220SKIPSKIPselectionSKIPplan-skip
1230SKIPSKIPselectionSKIPplan-skip
1240SKIPSKIPselectionSKIPplan-skip
1250SKIPSKIPselectionSKIPplan-skip
1260SKIPSKIPselectionSKIPplan-skip
1270SKIPSKIPselectionSKIPplan-skip
1280SKIPSKIPselectionSKIPplan-skip
1290SKIPSKIPselectionSKIPplan-skip
1300SKIPSKIPselectionSKIPplan-skip
1310SKIPSKIPselectionSKIPplan-skip
1320SKIPSKIPselectionSKIPplan-skip
1330SKIPSKIPselectionSKIPplan-skip
650CACHECACHEduplicateCONTINUEduplicate
650CACHECACHEpublic-routeCONTINUEpublic-channel
650CACHECACHEpublic-routeFREEpublic-channel
650CACHECACHEduplicateCONTINUEduplicate
650CACHECACHEsufficiencyCONTINUEsufficiency
Run context0——expansionCONTINUEexpansion
Run context1——attentionCONTINUEattention
Run context1——recommendationCONTINUEengine-recommendation
Run context1——recommendationCONTINUEengine-recommendation
981BUYSKIPduplicateCONTINUEduplicate
981BUYSKIPdiscussionCONTINUEdiscussion
981BUYSKIPattentionCONTINUEattention
981BUYSKIPpublic-routeCONTINUEpublic-channel
981BUYSKIPpublic-routeFREEpublic-channel
981BUYSKIPduplicateCONTINUEduplicate
151BUYSKIPduplicateCONTINUEduplicate
151BUYSKIPdiscussionCONTINUEdiscussion
151BUYSKIPattentionCONTINUEattention
151BUYSKIPpublic-routeCONTINUEpublic-channel
151BUYSKIPpublic-routeFREEpublic-channel
151BUYSKIPduplicateCONTINUEduplicate

Try a different limit

Compare only recorded budget and attention checks. Later reads, model choices and a whole alternative answer are unknown. These controls never change a live policy or authorize a purchase.

Portable research receipt

Take the evidence trail with you

One deterministic JSON bundle binds the answer, visible decisions, exact article versions, claim evidence and a Circle-settlement snapshot under SHA-256. Retain the digest to detect later changes; the self-check is not a publisher or Keryx signature.

View JSON ↗Download
Compare two saved reports
Manage question privacy

Recorded purchase outcomes

Recorded purchase outcomes are unavailable for this retained snapshot. No score or zero-cost success is inferred; the original report and receipt remain available.

Ask a follow-upNew dispatch · creators paid again

Carries this dispatch’s question as context — never its answer. The next dispatch is read from sources bought for it.